Google Ads
Brand versus non-brand: the split that keeps ads honest
Blended ROAS hides whether ads are finding new demand. Define brand once, split every channel by it, and report the two halves separately.
By CartKernel · Published
A single account-level return figure cannot tell you whether advertising is growing the business, because it mixes two activities with opposite economics. Serving an ad to somebody who typed your store name is order capture. Serving an ad to somebody searching for a product category is demand capture, and it costs several times more per sale. Averaged together, a rising blended number can mean the store got more famous, or that a campaign type quietly moved its spend onto brand queries.
The fix is a split that runs through every paid channel and through organic, applied the same way every month.
Define brand once, in writing
The definition has to be a rule anyone can apply, not a judgment call. Write it down and version it.
A brand query is one that contains:
- The store name, in any capitalization or spacing.
- Common misspellings and phonetic variants of the store name.
- The store name combined with anything else: reviews, discount code, shipping, a product line.
- Own-label product line names that exist only in your catalog.
Everything else is non-brand, including the manufacturer brands you resell. That last exclusion surprises people and it is the right call: somebody searching for a manufacturer’s product name has not chosen your store, so winning that click is acquisition even though a brand word appears in it. If you stock a manufacturer exclusively in your market, note the exception explicitly rather than blurring the rule.
Store the definition as a list and a pattern, keep it in the same place as the account naming convention, and update it when a new own-label line launches.
Split every channel, not just search
The split is only useful when it covers the whole picture.
| Channel | How the split is made |
|---|---|
| Search campaigns | Campaign structure, verified against the search terms report |
| Shopping | Campaign or product group structure, with brand negatives applied outside the brand tier |
| Broad-coverage campaign types | Brand exclusion controls plus post-hoc classification of the search terms that are reported |
| Organic search | Search Console queries classified with the same pattern |
| Everything else | Treated as non-brand by default, since it is not query-driven |
The organic row is worth keeping. Paid and organic brand demand rise and fall together, and a rise in paid brand conversions during a month when organic brand impressions also rose is a story about demand, not about the ad account.
Classifying spend where search terms are limited
Not every campaign type exposes a query for every click. Where that is the case, use the controls the campaign offers to exclude brand traffic in the first place, then classify what remains as non-brand and note the assumption in the report. Two safeguards keep this honest:
- Apply brand exclusions or account-level negative lists to any campaign whose job is acquisition, so brand traffic is prevented rather than subtracted afterward.
- Watch the shape of the numbers. A campaign whose reported return jumps sharply with no change to bids, budget or creative is usually serving new query types. Performance Max spending on brand covers how that shows up and what to do.
Where classification is genuinely impossible, report the campaign in a third column labeled unclassified rather than assigning it to a side. An honest unknown is more useful than a confident guess.
The report
One table, monthly, with a rolling thirteen-month history:
| Column | Why it is there |
|---|---|
| Spend, brand and non-brand | The actual split of the budget |
| Revenue, by the same split | Platform-reported, from a single consistent source |
| Return on ad spend, by split | Two numbers that mean two different things, never averaged |
| New customer share of orders, by split | Non-brand should carry a much higher share |
| Cost per new customer, non-brand | The number that answers whether acquisition is affordable |
| Brand impression share | Whether cheap, high-intent demand is being left uncovered |
| Total store revenue and total marketing spend | The denominator for blended efficiency |
Report platform figures and store figures side by side rather than trying to reconcile them into one truth. The reasons they differ are structural, and what attribution can and cannot tell you covers which of them are worth chasing.
What the split lets you decide
Four decisions become possible once the two halves are separated:
- How much acquisition actually costs. Non-brand cost per new customer, compared against contribution margin and repeat behavior, is the number that sets the budget. Blended CAC is the sanity check against the whole marketing spend.
- Whether growth is real. Revenue growth carried entirely by brand is growth created somewhere else, and the ad account is capturing it rather than causing it.
- Where the next dollar goes. A brand campaign below full coverage is usually the cheapest available conversion. Beyond full coverage, additional brand budget buys nothing.
- What to test for incrementality. The brand half is the classic candidate for a holdout, because much of its volume may arrive anyway. How do you measure incrementality for ecommerce ads covers the test designs.
Traps that corrupt the split
- Brand terms that contain category words. A store whose name includes a product noun will capture category queries in the brand bucket unless the pattern is written carefully. Test the pattern against a month of search terms before trusting it.
- Misspellings drifting. New misspellings appear as the brand grows. Refresh the list from the search terms report quarterly.
- Own-label lines missed. A new house brand launched without being added to the pattern shows up as non-brand and flatters acquisition performance.
- Retailer and marketplace names. Queries pairing your product with a marketplace are non-brand for this purpose, since the shopper is choosing a place to buy.
- Changing the definition mid-year. Any change breaks the trend. Version it, note the change date on the chart, and keep the old series.
Pair it with new versus returning revenue
The brand split describes intent. It does not, by itself, describe whether a customer is new. Run both, because a brand click from a first-time buyer and a brand click from a repeat customer have different value to the business. The measurement is covered in how do you track new versus returning customer revenue, and the bidding response is in bidding for new customers instead of cheap conversions.
Cadence and ownership
Run the split monthly for decisions and weekly for anomalies. The weekly check is short: did brand spend move more than a few percent, did non-brand cost per new customer step outside its usual band, did any campaign’s reported return jump without a change. The monthly review compares against the rolling history and sets the next month’s budget.
One person owns the definition, one report is the reference, and the MER calculator sits on top as the whole-business check. The reporting build itself is part of attribution reporting, and the account structure that makes the split possible without manual tagging is in Google Ads account structure for ecommerce.